feat(elasticache): open one port, switch to valkey

[hannahhoward]

Goals

Noticing high connect times for redis in traces -- probably just first query but still https://docs.aws.amazon.com/AmazonElastiCache/latest/dg/wwe-troubleshooting.html recommends having both ports open in case a client tries to connect to both

Also, we need to switch to valkey before we turn it on cause it's almost 1/3 cheaper per unit.

Implementation

• open both ports
• upgrade AWS provider to get valkey support
• move to valkey
• set a dummy password on the disabled default user -- no-password-required apparently not supported for valkey
• fix some downstream effects of create_before_destroy being applied at top of stack (shouldn't be as create_before_destroy will not work for many resources)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

see 1 file with indirect coverage changes

[volmedo]

tested this in my staging env and everything looks good, merging.

PS: I see a similar pattern of the first query to the cache being slow, it looks like it's due to warm up indeed.